ParasiticPattern
plain-language theorem explainer
A parasitic pattern is a pair of strictly positive reals: local reward and exported harm. Recognition Science treats this as the signature of evil: local stability bought by dumping cost elsewhere. The RS→RL bridge uses it to score and flag reward-hacking policies with unpriced externalities. The declaration is a pure data structure with positivity witnesses; no proof content.
Claim. A parasitic pattern is a pair $(r,h)\in\mathbb{R}_{>0}\times\mathbb{R}_{>0}$, where $r$ is local reward and $h$ is exported harm. The associated parasitism score is the ratio $h/r$; large values mark parasitic (evil) behavior under the RS→RL bridge.
background
The RS→RL bridge module casts Recognition Science as control theory for reinforcement learning: moral states with ledger/bonds/skew/energy, fourteen virtue generators as admissible actions, $\sigma=0$ feasibility via LACompletion, and Gibbs policies $p(a|s)\propto\exp(-J(s,a)/T_R)$ over undiscounted eight-tick windows (T6/T7).
In that setting, standard RL fails when rewards omit externalities: an agent can raise local return while dumping entropy or harm off-ledger. RS names that failure mode parasitism. The structure packages the two observables needed to detect it: a positive local reward and a positive exported harm (exported $\Delta S$ in the doc-comment).
Upstream ratio and score machinery (CPT ratio coordinates, CQ scores) supply the general pattern of comparing two positive rates; here the comparison is specialized to harm versus reward for the ethics audit layer.
proof idea
No proof. The declaration is a structure with four fields: two real coordinates and two positivity hypotheses $0 < \mathrm{localReward}$ and $0 < \mathrm{exportedHarm}$. Downstream, parasitismScore is the plain quotient of those coordinates, and isParasitic thresholds that quotient.
why it matters
This is the data carrier for the evil detector in the RS→RL stability audit. Downstream, parasitismScore forms the ratio exported-harm / local-reward, and isParasitic flags patterns above a threshold set at $1/\varphi$ (the RS $\varphi$-fraction bound). That closes the loop from the module's multi-objective selector (feasible → harm-minimax → value → robustness) to a concrete predicate against reward hacking.
Framework-wise it operationalizes the ethics side of Recognition Composition and ledger conservation: local $J$-gain that is not paid on-ledger appears as exported $\Delta S$. It does not itself invoke T5–T8, but it sits under the same eight-tick evaluation cadence and virtue-generator action basis that those forcing steps justify.
Switch to Lean above to see the machine-checked source, dependencies, and usage graph.